1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 */
20
21 /* This file was automatically generated --- DO NOT EDIT */
22 /* Generated on Thu Dec 10 07:04:41 EST 2020 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 14 -name n1fv_14 -include dft/simd/n1f.h */
29
30 /*
31 * This function contains 74 FP additions, 48 FP multiplications,
32 * (or, 32 additions, 6 multiplications, 42 fused multiply/add),
33 * 51 stack variables, 6 constants, and 28 memory accesses
34 */
35 #include "dft/simd/n1f.h"
36
n1fv_14(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1fv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
40 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
41 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
42 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
43 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
45 {
46 INT i;
47 const R *xi;
48 R *xo;
49 xi = ri;
50 xo = ro;
51 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
52 V T3, TH, Ts, TV, TW, Tt, Tu, TU, Ta, To, Th, Tp, TC, Tx, TK;
53 V TQ, TN, TR, T14, TZ, T1, T2;
54 T1 = LD(&(xi[0]), ivs, &(xi[0]));
55 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
56 T3 = VSUB(T1, T2);
57 TH = VADD(T1, T2);
58 {
59 V T6, TI, T9, TJ, Tn, TP, Tk, TO, Tg, TM, Td, TL;
60 {
61 V T4, T5, Ti, Tj;
62 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
63 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
64 T6 = VSUB(T4, T5);
65 TI = VADD(T4, T5);
66 {
67 V T7, T8, Tl, Tm;
68 T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
69 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
70 T9 = VSUB(T7, T8);
71 TJ = VADD(T7, T8);
72 Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
73 Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
74 Tn = VSUB(Tl, Tm);
75 TP = VADD(Tl, Tm);
76 }
77 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
78 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
79 Tk = VSUB(Ti, Tj);
80 TO = VADD(Ti, Tj);
81 {
82 V Te, Tf, Tb, Tc;
83 Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
84 Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
85 Tg = VSUB(Te, Tf);
86 TM = VADD(Te, Tf);
87 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
88 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
89 Td = VSUB(Tb, Tc);
90 TL = VADD(Tb, Tc);
91 }
92 }
93 Ts = VSUB(T9, T6);
94 TV = VSUB(TL, TM);
95 TW = VSUB(TJ, TI);
96 Tt = VSUB(Tn, Tk);
97 Tu = VSUB(Tg, Td);
98 TU = VSUB(TO, TP);
99 Ta = VADD(T6, T9);
100 To = VADD(Tk, Tn);
101 Th = VADD(Td, Tg);
102 Tp = VFNMS(LDK(KP356895867), Ta, To);
103 TC = VFNMS(LDK(KP356895867), To, Th);
104 Tx = VFNMS(LDK(KP356895867), Th, Ta);
105 TK = VADD(TI, TJ);
106 TQ = VADD(TO, TP);
107 TN = VADD(TL, TM);
108 TR = VFNMS(LDK(KP356895867), TQ, TN);
109 T14 = VFNMS(LDK(KP356895867), TN, TK);
110 TZ = VFNMS(LDK(KP356895867), TK, TQ);
111 }
112 ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)]));
113 ST(&(xo[0]), VADD(TH, VADD(TK, VADD(TN, TQ))), ovs, &(xo[0]));
114 {
115 V Tr, Tw, Tq, Tv;
116 Tq = VFNMS(LDK(KP692021471), Tp, Th);
117 Tr = VFNMS(LDK(KP900968867), Tq, T3);
118 Tv = VFMA(LDK(KP554958132), Tu, Tt);
119 Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts));
120 ST(&(xo[WS(os, 5)]), VFNMSI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
121 ST(&(xo[WS(os, 9)]), VFMAI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
122 }
123 {
124 V T16, T18, T15, T17;
125 T15 = VFNMS(LDK(KP692021471), T14, TQ);
126 T16 = VFNMS(LDK(KP900968867), T15, TH);
127 T17 = VFNMS(LDK(KP554958132), TU, TW);
128 T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV));
129 ST(&(xo[WS(os, 6)]), VFMAI(T18, T16), ovs, &(xo[0]));
130 ST(&(xo[WS(os, 8)]), VFNMSI(T18, T16), ovs, &(xo[0]));
131 }
132 {
133 V Tz, TB, Ty, TA;
134 Ty = VFNMS(LDK(KP692021471), Tx, To);
135 Tz = VFNMS(LDK(KP900968867), Ty, T3);
136 TA = VFMA(LDK(KP554958132), Tt, Ts);
137 TB = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TA, Tu));
138 ST(&(xo[WS(os, 13)]), VFNMSI(TB, Tz), ovs, &(xo[WS(os, 1)]));
139 ST(&(xo[WS(os, 1)]), VFMAI(TB, Tz), ovs, &(xo[WS(os, 1)]));
140 }
141 {
142 V TT, TY, TS, TX;
143 TS = VFNMS(LDK(KP692021471), TR, TK);
144 TT = VFNMS(LDK(KP900968867), TS, TH);
145 TX = VFMA(LDK(KP554958132), TW, TV);
146 TY = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TX, TU));
147 ST(&(xo[WS(os, 4)]), VFMAI(TY, TT), ovs, &(xo[0]));
148 ST(&(xo[WS(os, 10)]), VFNMSI(TY, TT), ovs, &(xo[0]));
149 }
150 {
151 V T11, T13, T10, T12;
152 T10 = VFNMS(LDK(KP692021471), TZ, TN);
153 T11 = VFNMS(LDK(KP900968867), T10, TH);
154 T12 = VFMA(LDK(KP554958132), TV, TU);
155 T13 = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), T12, TW));
156 ST(&(xo[WS(os, 2)]), VFMAI(T13, T11), ovs, &(xo[0]));
157 ST(&(xo[WS(os, 12)]), VFNMSI(T13, T11), ovs, &(xo[0]));
158 }
159 {
160 V TE, TG, TD, TF;
161 TD = VFNMS(LDK(KP692021471), TC, Ta);
162 TE = VFNMS(LDK(KP900968867), TD, T3);
163 TF = VFNMS(LDK(KP554958132), Ts, Tu);
164 TG = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TF, Tt));
165 ST(&(xo[WS(os, 11)]), VFNMSI(TG, TE), ovs, &(xo[WS(os, 1)]));
166 ST(&(xo[WS(os, 3)]), VFMAI(TG, TE), ovs, &(xo[WS(os, 1)]));
167 }
168 }
169 }
170 VLEAVE();
171 }
172
173 static const kdft_desc desc = { 14, XSIMD_STRING("n1fv_14"), { 32, 6, 42, 0 }, &GENUS, 0, 0, 0, 0 };
174
XSIMD(codelet_n1fv_14)175 void XSIMD(codelet_n1fv_14) (planner *p) { X(kdft_register) (p, n1fv_14, &desc);
176 }
177
178 #else
179
180 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 14 -name n1fv_14 -include dft/simd/n1f.h */
181
182 /*
183 * This function contains 74 FP additions, 36 FP multiplications,
184 * (or, 50 additions, 12 multiplications, 24 fused multiply/add),
185 * 33 stack variables, 6 constants, and 28 memory accesses
186 */
187 #include "dft/simd/n1f.h"
188
n1fv_14(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)189 static void n1fv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
190 {
191 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
192 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
193 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
194 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
195 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
196 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
197 {
198 INT i;
199 const R *xi;
200 R *xo;
201 xi = ri;
202 xo = ro;
203 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
204 V T3, Ty, To, TK, Tr, TE, Ta, TJ, Tq, TB, Th, TL, Ts, TH, T1;
205 V T2;
206 T1 = LD(&(xi[0]), ivs, &(xi[0]));
207 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
208 T3 = VSUB(T1, T2);
209 Ty = VADD(T1, T2);
210 {
211 V Tk, TC, Tn, TD;
212 {
213 V Ti, Tj, Tl, Tm;
214 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
215 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
216 Tk = VSUB(Ti, Tj);
217 TC = VADD(Ti, Tj);
218 Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
219 Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
220 Tn = VSUB(Tl, Tm);
221 TD = VADD(Tl, Tm);
222 }
223 To = VADD(Tk, Tn);
224 TK = VSUB(TC, TD);
225 Tr = VSUB(Tn, Tk);
226 TE = VADD(TC, TD);
227 }
228 {
229 V T6, Tz, T9, TA;
230 {
231 V T4, T5, T7, T8;
232 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
233 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
234 T6 = VSUB(T4, T5);
235 Tz = VADD(T4, T5);
236 T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
237 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
238 T9 = VSUB(T7, T8);
239 TA = VADD(T7, T8);
240 }
241 Ta = VADD(T6, T9);
242 TJ = VSUB(TA, Tz);
243 Tq = VSUB(T9, T6);
244 TB = VADD(Tz, TA);
245 }
246 {
247 V Td, TF, Tg, TG;
248 {
249 V Tb, Tc, Te, Tf;
250 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
251 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
252 Td = VSUB(Tb, Tc);
253 TF = VADD(Tb, Tc);
254 Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
255 Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
256 Tg = VSUB(Te, Tf);
257 TG = VADD(Te, Tf);
258 }
259 Th = VADD(Td, Tg);
260 TL = VSUB(TF, TG);
261 Ts = VSUB(Tg, Td);
262 TH = VADD(TF, TG);
263 }
264 ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)]));
265 ST(&(xo[0]), VADD(Ty, VADD(TB, VADD(TH, TE))), ovs, &(xo[0]));
266 {
267 V Tt, Tp, TP, TQ;
268 Tt = VBYI(VFNMS(LDK(KP781831482), Tr, VFNMS(LDK(KP433883739), Ts, VMUL(LDK(KP974927912), Tq))));
269 Tp = VFMA(LDK(KP623489801), To, VFNMS(LDK(KP900968867), Th, VFNMS(LDK(KP222520933), Ta, T3)));
270 ST(&(xo[WS(os, 5)]), VSUB(Tp, Tt), ovs, &(xo[WS(os, 1)]));
271 ST(&(xo[WS(os, 9)]), VADD(Tp, Tt), ovs, &(xo[WS(os, 1)]));
272 TP = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK))));
273 TQ = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TB, Ty)));
274 ST(&(xo[WS(os, 2)]), VADD(TP, TQ), ovs, &(xo[0]));
275 ST(&(xo[WS(os, 12)]), VSUB(TQ, TP), ovs, &(xo[0]));
276 }
277 {
278 V Tv, Tu, TM, TI;
279 Tv = VBYI(VFMA(LDK(KP781831482), Tq, VFMA(LDK(KP974927912), Ts, VMUL(LDK(KP433883739), Tr))));
280 Tu = VFMA(LDK(KP623489801), Ta, VFNMS(LDK(KP900968867), To, VFNMS(LDK(KP222520933), Th, T3)));
281 ST(&(xo[WS(os, 13)]), VSUB(Tu, Tv), ovs, &(xo[WS(os, 1)]));
282 ST(&(xo[WS(os, 1)]), VADD(Tu, Tv), ovs, &(xo[WS(os, 1)]));
283 TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ))));
284 TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TH, Ty)));
285 ST(&(xo[WS(os, 6)]), VSUB(TI, TM), ovs, &(xo[0]));
286 ST(&(xo[WS(os, 8)]), VADD(TM, TI), ovs, &(xo[0]));
287 }
288 {
289 V TO, TN, Tx, Tw;
290 TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL))));
291 TN = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP222520933), TE, VFNMS(LDK(KP900968867), TB, Ty)));
292 ST(&(xo[WS(os, 4)]), VSUB(TN, TO), ovs, &(xo[0]));
293 ST(&(xo[WS(os, 10)]), VADD(TO, TN), ovs, &(xo[0]));
294 Tx = VBYI(VFMA(LDK(KP433883739), Tq, VFNMS(LDK(KP781831482), Ts, VMUL(LDK(KP974927912), Tr))));
295 Tw = VFMA(LDK(KP623489801), Th, VFNMS(LDK(KP222520933), To, VFNMS(LDK(KP900968867), Ta, T3)));
296 ST(&(xo[WS(os, 11)]), VSUB(Tw, Tx), ovs, &(xo[WS(os, 1)]));
297 ST(&(xo[WS(os, 3)]), VADD(Tw, Tx), ovs, &(xo[WS(os, 1)]));
298 }
299 }
300 }
301 VLEAVE();
302 }
303
304 static const kdft_desc desc = { 14, XSIMD_STRING("n1fv_14"), { 50, 12, 24, 0 }, &GENUS, 0, 0, 0, 0 };
305
XSIMD(codelet_n1fv_14)306 void XSIMD(codelet_n1fv_14) (planner *p) { X(kdft_register) (p, n1fv_14, &desc);
307 }
308
309 #endif
310